Thin Film X-ray Reflectometry

Several 2008 accomplishments have enabled significant progress in the NIST pre-standards effort. We will be circulating our first XRR round robin study in 2009.

First, we have completed construction and assembly of a new goniometer base for the Ceramics Division Parallel Beam Diffractometer (CDPBD), NIST’s SI traceable X-ray reflectometer.Our NIST-designed and-constructed spherical air-bearing base can be seen in the figure below. This bearing allows us to completely separate sample translation and tilt, facilitating high accuracy XRR measurements.




Second, the CDPBD has been reconstructed on a granite base, further enhancing structural stability for high resolution, high accuracy XRR measurements. The instrument is shown during an alignment study.

Third, we have partnered with the International SEMATECH Manufacturing Initiative (ISMI) to develop pre-standards for our round robin and, eventually, feedstock structure for an XRR SRM. In 2008, ISMI manufactured its first test structure based on a calibration-optimized bilayer developed at NIST.

The NIST structural approach is to use a SiO2 capping layer to protect a high electron density (high-atomic number, Z) layer that is deposited on a silicon wafer. The thickness of the high-Z layer will provide the SI-traceable quantity in the final SRM. The surface oxide will be treated as a variable in the XRR analysis to accommodate any surface contamination in the field. The requirements for both layers in the structure are: 1) smooth deposition processing (low roughness), 2) uniform lateral deposition across the entire wafer surface, 3) no interdiffusion or interface layers formed during deposition, 4) good shelf stability. Thermal SiO2 was chosen for the capping layer due to its high uniformity, ubiquitous availability, and excellent stability.




The high-Z layer poses a significant challenge in structure development. In the first round of depositions, we chose Atomic Layer Deposition (ALD) TiN as the high-Z structural layer. Unfortunately, the deposition process generated a ˜1-nm interface layer, presumably at the Si substrate interface. This additional layer creates difficulties for XRR modeling and would reduce the overall effectiveness of the structure for instrument calibration. The following figure shows a Genetic Algorithm (GA) refinement performed at NIST using ISMI-measured XRR data for the (TiN deposition) on the test structure. The measurements were taken while the wafer was still within the SVTC Fab, and before SiO2 cap deposition.




We then used a statistically-based, Monte Carlo (MC) technique to determine the thicknesses of the TiN and TiSi interface layers. The figure below shows the posterior probability distribution for the TiN (t1) and TiSi (t2) from the MC analysis. The 1nm TiSi layer precludes accurate modeling of the TiN layer and Si substrate roughness, making this material and/or process inadequate for the XRR SRM.

In 2008, we presented our project at the spring ISMI Advanced CD Metrology Advisory Group (AMAG) meeting, established our development partnership, and presented preliminary results from our collaboration on XRR data analysis at the Denver X-ray Conference.